>The development *newn > *newm is plausible, given *septm and *dek^m,
it's just unfalsifiable.
The trouble is, it would have to occur early to prevent *n > *r in PIE,
whereas the evidence that lead to the suggestion of final -n in the first
place is late, e.g. from Latin. Perhaps there is a way forward by
reconsidering *newn as < **new- (= "new") + some suffix.
Peter